Welcome to the

AI Incident Database

A 15-year-old boy had planned, using Artificial Intelligence, to carry out a shooting at his school in 2027 and kill his classmates. Thanks to a warning from the FBI, the plan was thwarted. His home in Quilmes was raided, items related to the potential attack were seized, and his parents were summoned to the Prosecutor's Office.

The investigation by the Federal Department of Investigations (DFI) of the Argentine Federal Police began on August 19th, after the FBI's Legal Attaché at the U.S. Embassy in Argentina linked an email user who had been in contact with the ChatGPT group to organize the massacre.

The report indicated the minor's intention to carry out a shooting and kill his classmates, initially planned for 2027, but also suggested the possibility of moving it forward.

In his conversations with the Artificial Intelligence, the teenager had expressed his interest in acquiring clothing and tactical gear to carry out the massacre.

Upon receiving the alert, the Anti-Terrorism Investigation Unit conducted fieldwork and, through social media, was able to identify the teenager, as well as establish his address in the Quilmes district, in the southern part of Greater Buenos Aires.

With all the evidence presented, the Juvenile Criminal Responsibility Prosecutor's Office Number 1, headed by Dr. Walter Bruno, requested a search warrant from the Juvenile Court of Guarantees Number 1 of the Quilmes Judicial Department, presided over by Dr. Miriam Susana Alcolcel, with Dr. Mariana Digiani serving as the sole secretary.

The Raid in Quilmes and the Seized Material

Agents went to the teenager's home, where they entered and found items that matched the plan the suspect intended to carry out.

As a result of the operation, two cell phones, two pairs of tactical gloves, a military camouflage bucket hat, a skull-patterned balaclava, a military camouflage ski mask, and black tactical goggles were seized.

In addition, authorities summoned the parents and the minor to the Juvenile Criminal Responsibility Prosecutor's Office.

According to police sources, the teenager's identity is being kept strictly confidential for legal reasons, while the seized material was turned over to the court for forensic analysis.

Yemeni militants sought to use Anthropic’s Claude AI model to develop guidance, control and navigation software for its ballistic missiles, the US company said.

In its intelligence threat report released on Thursday, Anthropic did not name the Iranian-backed Houthis, but it said it “identified a cell of threat actors based in northern Yemen” — much of which is controlled by the Houthi rebels — “running three weapons development programmes”.

The study highlights the growing risks of advanced AI falling into the wrong hands, as the technology continues to improve rapidly.

The same Anthropic r

Elon Musk is planning to launch Cybercabs, fully autonomous vehicles without the means for a human to control them, any day now. But a lawsuit that closed this year indicates the electric car company may not be ready for the launch.

In March 2024, Neima Benavides and Dillon Angulo [filed a lawsuit](https://www.plainsite.org/courts/florida-southern-district-court/neima-benavides-personal-representative-of-the-estate-of-naibel-benavides-leon-deceas

Management agencies and authorities are trying to get the images taken offline. D66 is requesting a debate. The Dutch Data Protection Authority is urging victims to go to the Public Prosecution Service.

Caretaker Minister of Justice Dilan Yesilgöz is also urging victims to file a report. 'Disgusting and, moreover, a criminal offense. Please file a report!' she responds on X. The Minister notes that victims are even better protected by the Sexual Offenses Act, which she is defending in the Senate on Tuesday. 'These kinds of vile practices ought to be tackled! If you are a victim and you are seeking help, you can turn to helpwanted.nl.'

Deepfake

D66 MP Joost Sneller says on X that he has requested a debate regarding these so-called deepfakes. 'These companies and creators must not shirk their responsibility, and people must be protected by removal as soon as possible.' Bewildered

Confronted with the explicit videos, most victims react with bewilderment. Many are going to the police to file a report, including Mayor Sharon Dijksma (Utrecht), former police spokesperson Ellie Lust, various singers and presenters, and parties such as BBB, the Party for the Animals, and D66. GroenLinks-PvdA and VVD are also investigating legal action because their politicians appear in the videos. "Making and distributing these kinds of images is vile, offensive, and moreover, a criminal offense," states the VVD. The House of Representatives speaks of 'disgusting practices' and supports MPs in filing their reports.

What kind of deepfake videos are involved?

---

So-called deepfakes create a false reality that damages lives. Using AI technology, uploaders superimpose the face of a famous person onto that of a porn actress while she is having sex with herself or with a man. Anyone deliberately searching for the material will soon end up on Platform S*, a site that attracts 13 million visitors monthly and has a huge database of videos featuring Dutch celebrities, ranging from artists, presenters from NOS, SBS6, and ESPN, to Olympic athletes, and from (former) ministers, mayors, and members of the Royal Family.

The group behind Platform S, founded in 2018, operates secretly. Uploaders are advised to 'always remain anonymous,' presumably to avoid prosecution. The better a video scores, the better they are paid. Some videos, with explicit titles such as 'Person A cums' or 'Person B gets fucked by,' have already been viewed more than 100,000 times. Below the footage, viewers encourage the uploaders and offer advice for future victims. 'You surely want more of this' and 'You should try (...) next!' can be read, among other things. Without government intervention, everyone will be targeted, cyber experts predict. Victims are receiving insufficient help because proper legislation is lacking, warns tech lawyer Jelle Wieringa. As a result, the number of videos featuring Dutch celebrities continues to rise. Just a few weeks ago, for instance, a new video featuring a former minister appeared online. "It is time this received major media coverage, because things are going from bad to worse," says SBS6 presenter Sylvana IJsselmuiden, one of the victims. Her video has already been viewed nearly 20,000 times. She is considering filing a police report. "I have already tried to get it taken down."

Her reaction is not an isolated case. SBS6 prominent Hélène Hendriks is also working behind the scenes to have the footage removed. The managers of Leontine Ruiters, SBS presenter Noa Vahle, actress Jennifer Hoffman, Olcay Gulsen, and singer Simone Simons are currently exploring their legal options as well. A former Eurovision Song Contest contestant and a member of the Senate say they will file a police report regardless, in the hope that the videos will be taken 'offline as soon as possible'. Management agency A Million Faces also calls it 'bizarre' that 'a number of our talents' appear in the videos and is considering further steps.

What can victims do?

---

Deepfake falls under the category of revenge porn, which is a criminal offense. Legally, there are several avenues to pursue. Victims can file a police report, which can lead to an investigation led by a public prosecutor. With the approval of the examining magistrate, the Public Prosecution Service can attempt to shut down the site. This is only possible if the criminal offenses involved warrant arrest, the Public Prosecution Service explains. Practically taking it offline depends on various factors, such as the location of the servers running the site.

That is difficult in the case of Platform S. The site is registered in Charlestown, the capital of the Caribbean island of Saint Kitts and Nevis, next to Sint Eustatius. Probably not without reason, because prosecution becomes extra difficult as a result, says IT expert Arnoud Engelfriet of consultancy firm ICTRecht. Another route is to track down the uploaders themselves. This is possible if the police issue an order for the surrender of IP addresses and Platform S complies with it. The latter, however, requires cooperation from the local police: difficult, therefore.


If I were a victim, I would write to the publisher, provider, and domain name holder and threaten them with hell and damnation

-Tech lawyer Jan Gerrit Kroon

Mass claim

Another option is to start a mass claim, such as a foundation where victims unite. Lawyer Richard Korver, chairman of the LANZGS foundation, which includes specialized victim advocates, indicates that victims would be wise to file a police report. Additionally, victims can file a complaint with the Dutch Data Protection Authority for violation of privacy. Finally, aggrieved parties can contact the hosting providers. "They have policies that oblige them to intervene in illegal activities on websites they host," explains tech lawyer Jan Gerrit Kroon.

Prosecution therefore certainly has a chance of success. For instance, last year the creator of the deepfake videos of WNL presenter Welmoed Sijtsma was sentenced to community service. That conviction did not go unnoticed among deepfake enthusiasts. One of the more popular uploaders jokingly calls himself 'Welmoed Sijtsma'. His profile features over 120 manipulated videos, in which a striking number of politicians are exploited. Sijtsma, who previously made a documentary series about the world of deepfakes, was unaware of this and calls it 'bizarre'.


There is currently no deepfake police, so victims really have to stand up for themselves

-AI expert Henk van Ess

'Threaten with hell and damnation'

Whichever path you choose: 'please' do not let it go, legal experts advise. "If I were a victim, I would write to the publisher, provider, and domain name holder and threaten them with hell and damnation," advises tech lawyer Jan Gerrit Kroon. There is currently no deepfake police, so victims really have to stand up for themselves, states AI expert Henk van Ess. "Victims must keep law enforcement and legislators on their toes and definitely continue filing reports."

The homemade videos on Platform S usually last a few minutes and sometimes link to even longer videos offered for a fee on another website—which we will call Platform B for convenience. It is noteworthy that, according to NBC, payments on that site are handled by a Dutch service provider named Verotel, based in Amsterdam. An English-speaking employee refuses to comment on the complaints experienced by Dutch celebrities and hangs up the phone.

A day after this publication, Verotel finally provides full disclosure via a press statement sent by email by CEO Joost Zuurbier. It states that the collaboration with Platform B was terminated in March 2023 because the platform (similar to OnlyFans) allegedly did not comply with the 'strict rules'. Verotel wishes to emphasize that it was never directly involved with Platform S. The Amsterdam-based service provider states that it strongly condemns the creation and distribution of deepfake videos.

'It is a fantasy'

Incidentally, there are already porn sites that actively ban deepfake from their platforms. Anyone searching for such material on Pornhub, for example, is told that the search term contains 'illegal and abusive sexual material' for which no consent has been given. Pornhub warns that distributing these types of videos carries sanctions. However, another popular site does feature hundreds of deepfake photos of over 200 Dutch celebrities, including models and popular YouTubers.

Meanwhile, the anonymous American owner of Platform S claims to be unaware of any wrongdoing. "As long as you tell yourself that it isn't all real, it doesn't matter that much," he says in a rare interview with the BBC. "I don't really feel that permission is required for this. It is a fantasy, it is not real." The owner did not respond to questions from this site.

*Disclaimer: For this article, dozens of victims and their spokespersons, organizations, political parties, and managers were interviewed. All parties involved reviewed the article in advance and agreed to its publication. The AD understands that mentioning names may be accompanied by undesirable effects. The goal of this article is to raise awareness and to bring criminal offenses to light. Platform S is not the correct name of the deepfake website mentioned. This was done deliberately to protect the victims.

Anthropic on ​Wednesday disclosed another instance of an AI model hacking external systems during testing, the latest in a ‌growing list of such incidents that have raised concerns about the risk posed by autonomous AI agents.

The January incident went undetected until last month, despite an earlier company-wide review, Anthropic said, underscoring the challenge that AI developers face in identifying and containing unexpected behavior by advanced models.

The company ​said in a blog post the incident involved an early version of Claude Opus 4.6. It said it ​had notified all the affected parti